A method and an arrangement for controlling the drive unit of a motor vehicle is known from U.S. Pat. No. 5,692,472. Here, a torque of the drive unit is controlled in dependence upon a desired value for this torque. To ensure the operational reliability of this control system, a maximum permissible torque of the drive unit is also formed and is compared to an actual torque of the drive unit and fault reaction measures are initiated when the actual torque of the drive unit exceeds the maximum permissible torque. In an embodiment described there, the maximum permissible torque is formed in dependence upon the desired torque value. This value, in turn, is computed on the basis of the position of an operator-controlled element, such as an accelerator pedal, which is operable by the driver, or is computed in dependence upon the desired torque pregiven by other control systems or control functions, for example, in dependence upon a desired torque of an engine drag torque control and/or of an idle rpm control. The maximum permissible torque is formed by means of a characteristic line or a characteristic field in dependence upon the desired torque value. There is no description of a consideration of the tolerances of the drive unit which, for example, are caused by internal friction. Furthermore, the permissible torque is dependent upon the driver command torque so that the permissible torque is likewise defective for a theoretically conceivable error in the computation of this torque.